🤖 AI Agent 研究Research
计算机使用代理（ CUA ）正在迅速改进，但大规模合成多样化和可靠的软件环境仍然具有挑战性。本机应用程序环境需要大量的构建和集成工作，这限制了应用程序的覆盖范围，并增加了特定于应用程序的过度拟合的风险。
Computer-use agents (CUAs) are rapidly improving, yet synthesizing diverse and reliable software environments at scale remains challenging. Native application environments require substantial construction and integration effort, which limits application coverage and increases the risk of application-specific overfitting.
编码代理的评估通常力求尽可能真实。在我们的研究中，我们收集了JetBrains IDE中真实软件工程师的4,782个代理会话，我们称之为生产会话。
The evaluation of coding agents generally strives to be as realistic as possible. In our study, we collect 4,782 agent sessions of real software engineers in JetBrains IDEs, which we call Production Sessions.
大型基础模型加速了通用代理的进展，这些通用代理通过语言和多模态信号与人类和其他代理进行交互。然而，强大的多代理决策需要推理其他代理在部分可观察性下知道、打算和可能做的事情。
Large foundation models have accelerated progress toward general-purpose agents that interact with humans and other agents through language and multimodal signals. However, robust multi-agent decision-making requires reasoning about what other agents know, intend, and are likely to do under partial observability.
对抗性游戏推动了从启发式搜索到强化学习的进步，但从有限的样本中学习和调整策略仍然具有挑战性。人工智能代理通过将游戏体验转化为可执行策略的修订，提供了另一种选择。
Adversarial games have driven advances from heuristic search to reinforcement learning, yet learning and adapting strategies from limited samples remain challenging. AI agents offer an alternative by turning game experience into revisions of executable policies.
⭐ GitHub 热门项目GitHub Trending
【GitHub】🎯 免费开源的人工智能面试助理。编码和求职面试的语音+屏幕截图实时解答。适用于Groq、Claude、Gemini、GPT、DeepSeek、Ollama。Clu的免费替代品（⭐ 4 ）
【GitHub】🎯 Free & open-source AI interview assistant. Real-time answers from voice + screenshots for coding and job interviews. Works with Groq, Claude, Gemini, GPT, DeepSeek, Ollama. A free alternative to Clu （⭐ 4）
【GitHub】用简单的语言在MICrONS小鼠视觉皮层EM数据集中查找细胞，并在Neuroglancer中从任何AI （ MCP服务器， CLI ， Python ）中打开它们。（⭐ 0 ）
【GitHub】Find cells in the MICrONS mouse visual cortex EM dataset in plain language and open them in Neuroglancer, from any AI (MCP server, CLI, Python). （⭐ 0）
【GitHub】在一行中描述任何机器。Claude在一个HTML文件中构建了一个基于物理学的真实交互式模拟，再加上要发布的影片。开源Claude Code技能。（⭐ 6 ）
【GitHub】Describe any machine in one line. Claude builds a photoreal, physics based, interactive simulation of it in one HTML file, plus the film to post it. An open source Claude Code skill. （⭐ 6）
【GitHub】Claude Code会话在重新启动后仍存活的开源终端，每个正在运行的代理只需单击一下即可。曲速叉（⭐ 1 ）
【GitHub】Open-source terminal where Claude Code sessions survive restarts and every running agent is one click away. A Warp fork. （⭐ 1）
🚀 模型与行业动态Models & Industry
Muse Spark的目标是成为一个可以在低延迟下执行的轻量级模型，而Wonder模型更大，据说更有能力，在编码等复杂任务上具有更高的基准分数和更好的结果。
Where Muse Spark aimed to be a lightweight model that could perform at low latency, the Wonder models are larger and said to be more capable, with higher benchmark scores and better results on complex tasks like coding.
微软总裁布拉德·史密斯（ Brad Smith ）呼吁对人工智能进行紧急监管，称牛奶和洗衣机比人工智能公司拥有更多的监
Microsoft president Brad Smith calls for urgent AI regulation, saying milk and washing machines have more oversight than AI companies.
Blueprints AI是一家总部位于圣莫尼卡的初创公司，它开发的软件可以从照片、PDF或草图中生成符合代码的施工蓝图。
Blueprints AI, a bootstrapped startup based in Santa Monica, developed software that produces code-compliant construction blueprints from a photo, PDF, or sketch.
人工智能的核心承诺之一是，它对一个人的了解越多，就越能更好地帮助他们完成任务并提出适当的建议，如果你相信科技公司的话，这需要应用程序和设备始终监听、观看或记录用户的活动，并收集所有可能的背景信息。
One of the core promises of AI is that the more it knows about a person, the better it will be able to assist them with tasks and make appropriate suggestions, which, if you'll believe the tech companies, requires apps and devices to listen, watch, or record users' activity all the time and gather all possible context.
🔥 社区热议Community
【Lobsters】热度: 48↑ | 66 评论 | 标签: linux, vibecoding
【Lobsters】热度: 48↑ | 66 评论 | 标签: linux, vibecoding
【HN】热度: 33 分 | 15 评论
【HN】热度: 33 分 | 15 评论
【HN】热度: 102 分 | 69 评论
【HN】热度: 102 分 | 69 评论